Sum-product exponent for the reals

Description of constant

For a finite set $A \subset \mathbb{R}$ write \(A+A = \\{ a+b : a,b \in A \\}, \qquad AA = \\{ ab : a,b \in A \\}\) for the sumset and product set. The (real) sum-product exponent is \(C_{84b} := \liminf_{n \to \infty}\ \min_{\substack{A \subset \mathbb{R} \\ \lvert A\rvert = n}} \frac{\log \max(\lvert A+A\rvert, \lvert AA\rvert)}{\log n}.\) Equivalently, $C_{84b}$ is the largest $s$ such that $\max(\lvert A+A\rvert, \lvert AA\rvert) \gg_\epsilon \lvert A\rvert^{s-\epsilon}$ for every $\epsilon > 0$ and every finite $A \subset \mathbb{R}$.

Erdős and Szemerédi [ErSz83] conjectured (originally for $A \subset \mathbb{Z}$, but also for $A \subset \mathbb{R}$) that $C_{84b} = 2$: at least one of the sumset or product set must be of near-maximal size $\lvert A\rvert^{2-o(1)}$. This sum-product conjecture for the reals was disproved in May 2026 by Bloom, Sawin, Schildkraut and Zhelezov [BSSZ2026], who showed $C_{84b} < 2$.

Known upper bounds

Bound Reference Comments
$2$ [ErSz83] Erdős–Szemerédi constructed sets of integers with $\max(\lvert A+A\rvert, \lvert AA\rvert) \le \lvert A\rvert^{2-c/\log\log\lvert A\rvert}$, so the exponent is $\le 2$. This was conjectured to be sharp (the sum-product conjecture).
$2 - c$ for an absolute $c>0$ [BSSZ2026] Bloom–Sawin–Schildkraut–Zhelezov disprove the conjecture by constructing arbitrarily large $A \subset \mathbb{R}$ — algebraic integers in totally real number fields of degree $\asymp \log\lvert A\rvert$ — with $\max(\lvert A+A\rvert, \lvert AA\rvert) \le \lvert A\rvert^{2-c}$. A non-optimized explicit version of the argument gives $c \ge 0.00000087$, i.e. exponent $\le 1.99999913$ (the authors stress this value “should not be taken too seriously”).
$1.999281$ [Al26], [EPF52] Althoefer (28 May 2026), posted on the Erdős Problems forum [EPF52]; a ChatGPT 5.5 long-thinking optimization of the explicit constant of [BSSZ2026, §5] giving $c \ge 0.000719$. Unverified.

Known lower bounds

Bound Reference Comments
$1$ Trivial $\lvert A+A\rvert \ge 2\lvert A\rvert - 1$.
$5/4 = 1.25$ [El97] Elekes, via the Szemerédi–Trotter incidence theorem.
$14/11 \approx 1.2727$ [So05] Solymosi.
$4/3 \approx 1.33333$ [So09] Solymosi; the long-standing “$4/3$ barrier”.
$\approx 1.33338$ [KoSh16] Konyagin–Shkredov; first to break the $4/3$ barrier (their 2015 and 2016 papers give $\approx 1.33338$ and $\approx 1.33384$).
$\approx 1.33428$ [Sh19] Shakan.
$\tfrac{1558}{1167} \approx 1.33504$ [RuSt22] Rudnev–Stevens.
$\tfrac{1270}{951} \approx 1.33543$ [Bl25] Bloom.
$\tfrac{4}{3} + \tfrac{10}{4407} \approx 1.335602$ [Cu25] Cushman; current record.

References

Contribution notes

Prepared with assistance from Claude Opus 4.7, which read the [BSSZ2026] preprint and Bloom’s history page to assemble the bounds. The pre-2026 lower-bound citations were compiled from Bloom’s history page together with the reference list of [BSSZ2026], and the exact published details (volumes, page numbers, the [Bl25] title) should be independently verified before citation.